Accuracy of this project .
Opened this issue · 3 comments
jiguanglu commented
Hi:
First THX for sharing this project.
I Iterated 200 times,and the result is :
**After 2000 epochs train_acc is 1.0000 and test_acc is 0.8480
Time usage: 0:28:03**
But when I test a new image use "prediction_online.py", I got a bad result.
['psi_8_5']
[0.87777895]
psi_8_5 0.87777895 ./static/latest_streaming_frame_annotated.jpg
Can you give me some sugesstion?
mekhod commented
Hi,
Thank you for your question. When we developed this model we got an
accuracy of around 0.94 on the test images, so based on the message you
have received you do not have the same model because your test accuracy is
0.84. I do not work at OCI (Object Computing Inc.) anymore so please ask
this question from their ML team and they will answer your question.
Thanks,
Mehdi Khodayari
…On Wed, Nov 13, 2019 at 11:10 PM lujiguang ***@***.***> wrote:
Hi:
First THX for sharing this project.
I Iterated 200 times,and the result is :
**After 2000 epochs train_acc is 1.0000 and test_acc is 0.8480
Time usage: 0:28:03**
But when I test a new image use "prediction_online.py", I got a bad result.
['psi_8_5']
[0.87777895]
psi_8_5 0.87777895 ./static/latest_streaming_frame_annotated.jpg
[image: gauge_scale_6]
<https://user-images.githubusercontent.com/16578290/68828226-e9e82400-06df-11ea-897b-67643307b7e9.jpg>
Can you give me some sugesstion?
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#21?email_source=notifications&email_token=AE26FFPQGM43AM5JF3KL77LQTTMSVA5CNFSM4JNF5K7KYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HZGVVAQ>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AE26FFPCPQCLQ73X2O4LWK3QTTMSVANCNFSM4JNF5K7A>
.
jiguanglu commented
@mekhod
THX for your replay, but I don't know how can I contact ML team?
MVG(from Germany)
Jiguang Lu
MuhammadAsadJaved commented
@jiguanglu I think you can try this email address.
I am also working on a similar problem. Have you resolved the issue or build a better solution?
I hope we can help each other.